目前國內學界雖然大量地採用以調查研究方法所蒐集的資料來進行研究分析、理論建構與假設驗證，卻鮮少注意到在以調查研究法蒐集資料過程中可能會發生的問題。本文從調查研究的方法論角度，檢視1986至1998年間台灣地區民眾參與訪問的態度及其變遷。本研究發現：隨著台灣的政治民主化與社會開放，民眾有愈來愈不願接受訪問的傾向。訪問失敗有四分之一左右是因為「接觸不到受訪者」、四分之一是「接觸但失敗」，而「拒訪」的比例有緩慢上升的現象。在「選舉」及「兩岸關係」主題的兩類訪問中，「兩岸關係」類主題的拒訪率高於「選舉」類主題。受訪者對敏感性問題的回答意願，也因為受訪者的背景以及本身所具備的資訊而出現差異。當我們以受訪者是否透露其投票對象當作測量對敏感性問題回答意願時發現：男性、年輕、外省籍、高教育程度以及具有政黨認同者，較傾向願意回答敏感性問題。選舉的類型對受訪者回答敏感性問題的意願也有影響。我們發現總統以及省市長選舉，因為選區幅員較大，參選人數少，使得選民不願意回答投票對象的比例降低許多。本文中，也對訪問參與這個主題的研究，提出若干建議。

In Taiwan, researchers and scholars widely use survey data to conduct research analysis, construct theories and test hypotheses. However, problems came from process of data collection seldom been noticed. From the perspective of survey research methodolo-gy, this paper examined people's attitudes on survey participation in Taiwan between 1986 and 1998. This research showed that, in Tai-wan, people became less likely to participate in survey interview even political environment became more democratic and open in the last decade. Among people who did not participate in survey, one quarter of them can not be reached by interviewers, and one out of four can be reached but refused to participate in survey. Worthy to be noticed is that the ratio of "refuse to be interviewed" increased gradually. Compared with survey on "election studies", sur-vey topic on "relation between Taiwan and Mainland China" re-ceived higher refuseal rate. The paper indicated that the response rate of sensitive questions was influenced by respondents' demo-graphic background and level of information they had. Men, youths, mainlanders, high-educated and party identifiers were more likely to answer the sensitive question-whom did you vote for- in this research. Type of elections also playd a role on re-sponse rate. In the survey on presidential election and gubernator-rial/mayoral election, nonresponse rates on reporting preferential candidates were lower than nonresponse rates in legislative elec-tions. Therefore, the difference of electoral system between sin-gle member district and SNTV might affect nonresponse rates. Several methodological suggestions are also presented for future sur-vey participation research.